Enhancing Arabic Dialectical Speech Recognition with Large Language Models
Islamic University Journal of Applied Sciences · 01.07.2026 · VIII (I) · 280-298
DOI: 10.63070/jesc.2026.028 ↗Kurz gefasst
Fachartikel in der Zeitschrift Islamic University Journal of Applied Sciences, online erschienen am 01.07.2026. Verfasst von Rehab Albeladi.
Abstract (Originalsprache)
Automatic Speech Recognition (ASR) for Arabic faces unique challenges due to the language’s diglossic nature, where Modern Standard Arabic (MSA) coexists with diverse dialects that differ substantially at lexical, phonological, and morphological levels. In this study, we present a comparative evaluation of state-of-the-art ASR models, Whisper (medium and large) and Wav2Vec2, applied to five Arabic varieties: Egyptian, Hijazi, Khaliji, Najdi, and MSA. We design a two-phase experiment: in the first phase, 50 speech samples per dialect are evaluated under three settings, raw transcriptions, and outputs corrected by two large language models (GPT and ALLAM). In the second phase, the best-performing ASR–LLM combination is scaled to the full test set for each dialect. Performance is assessed using Word Error Rate (WER) and semantic similarity distance. Results show that Whisper-Large combined with GPT consistently outperforms other systems across dialects. Improvements are especially pronounced in dialectal speech, while MSA achieves the lowest baseline WER. These findings highlight both the persistent difficulty of dialectal Arabic ASR and the effectiveness of LLM-based post-processing for reducing recognition errors.
Zitieren
APA
Albeladi, R. (2026). Enhancing Arabic Dialectical Speech Recognition with Large Language Models. Islamic University Journal of Applied Sciences, VIII(I), 280-298. https://doi.org/10.63070/jesc.2026.028
Vancouver
Albeladi R. Enhancing Arabic Dialectical Speech Recognition with Large Language Models. Islamic University Journal of Applied Sciences. 2026;VIII(I):280-298. doi:10.63070/jesc.2026.028
BibTeX
@article{Albeladi2026,
author = {Albeladi, Rehab},
title = {Enhancing Arabic Dialectical Speech Recognition with Large Language Models},
journal = {Islamic University Journal of Applied Sciences},
year = {2026},
volume = {VIII},
number = {I},
pages = {280--298},
doi = {10.63070/jesc.2026.028}
}
EUVIMED